Multiscale and probabilistic modelling of progressive slope failure

Lead Research Organisation: Brunel University
Department Name: Civil and Environmental Engineering


In the UK and globally, the slope failures of various sizes are crucially affecting the sustainable development of resilient cities, as its occurrence can significantly threaten the populations, infrastructures, public services, and environment. For example, the British Geological Survey has estimated that 10% of slopes in the UK are classified as at moderate to significant landslide risk, with more than 7% of the main transport networks located in these areas. These slopes may fail during prolonged periods of wet weather or more intensive short duration rainfall events.

To date, the public awareness of slope failure risk is high, but our understanding of its fundamental failure mechanism and countermeasures are still very limited. This is mainly due to the difficulties in analysing the multiscale responses and characterize the spatial inhomogeneity of material properties of slopes. Laboratory and numerical investigations with well-developed empirical models can explain the general features of some specific slope failure events but cannot be applied universally. Some challenging issues need to be addressed, such as i) How to develop reliable mathematical models with multiscale modelling capability to analyse the progressive failure of slopes? ii) How to address the spatial variabilities and uncertainties of real slopes, e.g. material property, fractures, fluid permeability? iii) How to accurately estimate the spreading of landslide and its impact on infrastructures? The fundamental scientific issue of these challenges is the weakening mechanism of inhomogeneous slopes at different scales as it determines the slope responses under various geological and environmental conditions.

The proposed research aims to explore the fundamental mechanism of progressive slope failure and its impacts on infrastructures via a multiscale and probabilistic modelling approach. It enables the large deformation of slopes to be conveniently analysed by FEM as boundary value problem (BVP), while the local fracturing, cracking, or discontinuous behaviours of soil to be evaluated in smaller discrete subdomains through granular mechanics by DEM. The boundary condition of DEM assembly is derived from the global deformation of FEM meshes. In the analysis, the soil/rock properties (e.g. elastic modulus, friction coefficient, strength, and fluid permeability) will be evaluated as random fields with spatial variabilities. The numerical modelling can effectively bridge the gap between the microscopic material properties and the overall macroscopic slope responses. In the numerical modelling, the contributions of material inhomogeneity and discontinuity to slope failure and subsequence landslide spreading can be effectively investigated. The internal fracture would occur naturally when the loading stress exceeds the particle bonding strength at the microscale, which avoids the use of some phenomenological constitutive laws in conventional continuum modelling.

As a multidisciplinary research, this project will involve the subjects of geotechnical engineering, computational geotechnics, geology, statistics, soil/rock mechanics and granular mechanics. The proposed numerical model will benefit all researchers and stakeholders in land planning and management by providing efficient and reliable numerical modelling approaches. This will support the landslide risk evaluation, hazard mitigation and long-term land management, from which the environmental, social, and economic benefits can be achieved. As a result, the decision makers would have greater confidence in slope failure risk assessments on which they are basing their infrastructure investment considerations. Consequently, hazard warning systems, protections and land utilization regulations can be implemented, so that the loss of lives and properties can be minimized without investing in long-term, costly projects of ground stabilization.


10 25 50